European Journal of Combinatorics 31 (2010) 1714–1724

Contents lists available at ScienceDirect

European Journal of Combinatorics

journal homepage: www.elsevier.com/locate/ejc

Characterization of graphs using domination polynomials

Saieed Akbari a,b , Saeid Alikhani c , Yee-hock Peng d,e
a
Department of Mathematical Sciences, Sharif University of Technology, Tehran, Iran
b
School of Mathematics, Institute for Research in Fundamental Sciences (IPM), Tehran, Iran
c
Department of Mathematics, Yazd University, 89175-741, Yazd, Iran
d
Department of Mathematics, University Putra Malaysia, 43400 UPM Serdang, Malaysia
e
Institute for Mathematical Research, University Putra Malaysia, 43400 UPM Serdang, Malaysia

article info abstract

Article history: Let G be a simple graph of orderPn. The domination polynomial of
G is the polynomial D(G, x) = i=1 d(G, i)x , where d(G, i) is the
Received 13 December 2008 n i
Received in revised form number of dominating sets of G of size i. A root of D(G, x) is called a
31 January 2010
domination root of G. We denote the set of distinct domination roots
Accepted 12 March 2010
Available online 22 April 2010
by Z (D(G, x)). Two graphs G and H are said to be D -equivalent,
written as G ∼ H, if D(G, x) = D(H , x). The D -equivalence class
of G is [G] = {H : H ∼ G}. A graph G is said to be D -unique
if [G] = {G}. In this paper, we show that if a graph G has two
distinct domination roots, then Z (D(G, x)) = {−2, 0}. Also, if G is
a graph with no pendant vertex and has three √distinct domination
√
roots, then Z (D(G, x)) ⊆ {0, −2 ± 2i, −3±2 3i }. Also, we study
the D -equivalence classes of some certain graphs. It is shown that
if n ≡ 0, 2 (mod3), then Cn is D -unique, and if n ≡ 0 (mod 3), then
[Pn ] consists of exactly two graphs.
© 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Let G = (V , E ) be a simple graph. The order of G is the number of vertices of G. For any vertex
v ∈ V , the open neighborhood of v is the set N (v) = {u ∈ V |uv ∈ E } and the closed neighborhood is
the set N [v] = N (v) ∪ {v}. For a set S ⊆ V , the open neighborhood of S is N (S ) = v∈S N (v) and
S
the closed neighborhood of S is N [S ] = N (S ) ∪ S. A set S ⊆ V is a dominating set if N [S ] = V , or
equivalently, every vertex in V \ S is adjacent to at least one vertex in S. The domination number γ (G)
is the minimum cardinality of a dominating set in G. A dominating set with cardinality γ (G) is called
a γ -set. For a detailed treatment of this parameter, the reader is referred to [5]. An i-subset of V (G) is a

E-mail address: s_akbari@sharif.edu (S. Akbari).

0195-6698/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ejc.2010.03.007
 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724 1715

subset of V (G) of cardinality i. Let D (G, i) be the family of dominating sets of G which are i-subsets and
P|V (G)|
let d(G, i) = |D (G, i)|. The polynomial D(G, x) = i=1 d(G, i)xi is defined as domination polynomial
of G [1]. A root of D(G, x) is called a domination root of G. In this paper, the set of distinct roots of
D(G, x) is denoted by Z (D(G, x)).
The corona of two graphs G1 and G2 , as defined by Frucht and Harary in [4], is the graph G = G1 ◦ G2
formed from one copy of G1 and |V (G1 )| copies of G2 , where the ith vertex of G1 is adjacent to every
vertex in the ith copy of G2 . The corona G ◦ K1 , in particular, is the graph constructed from a copy of
G, where for each vertex v ∈ V (G), a new vertex v 0 and a pendant edge vv 0 are added. The join of
two graphs G1 and G2 , denoted by G1 ∨ G2 , is a graph with vertex set V (G1 ) ∪ V (G2 ) and edge set
E (G1 ) ∪ E (G2 ) ∪ {uv|u ∈ V (G1 ) and v ∈ V (G2 )}.
Non-isomorphic graphs may have the same domination polynomial. Two graphs G and H are said to
be dominating equivalent, or simply D -equivalent, written as G ∼ H, if D(G, x) = D(H , x). It is evident
that the relation ∼ of being D -equivalent is an equivalence relation on the family G of graphs, and
thus G is partitioned into equivalence classes, called the D -equivalence classes. Given G ∈ G, let
[G] = {H ∈ G : H ∼ G}.
We call [G] the equivalence class determined by G. A graph G is said to be dominating unique or, simply
D -unique, if [G] = {G}. There are two interesting problems on equivalence classes:
(i) Which graphs are D -unique?
(ii) Determine the D -equivalence classes for some families of graphs.
As usual we denote the complete graph, path and cycle of order n by Kn , Pn and Cn , respectively. Also
K1,n is the star graph with n + 1 vertices. For a graph G, a connected component isomorphic to K2 , is
called a K2 -component.
In the next section, we study the domination polynomial of G ◦ K1 , for any graph G, and we
characterize the class [G ◦ K1 ]. In Section 3, we characterize all graphs with at most two domination
roots, and in Section 4, we do the same for all graphs with exactly three domination roots. Finally, in
the last section we study the D -equivalence classes of certain graphs.

2. The domination polynomial of G ◦ K1

Let G be any graph with vertex set {v1 , . . . , vn }. Add n new vertices {u1 , . . . , un } and join ui to vi for
every i, 1 ≤ i ≤ n. By definition this graph is G ◦ K1 . We start this section with the following lemma.

Lemma 1. For any graph G of order n, γ (G ◦ K1 ) = n.

Proof. If D is a dominating set of G, then for every i, 1 ≤ i ≤ n, ui ∈ D or vi ∈ D. This implies that
|D| ≥ n. Since {u1 , . . . , un } is a dominating set of G ◦ K1 , we have the result. 
By Lemma 1, d(G ◦ K1 , m) = 0, for every m, m < n, so we shall compute d(G ◦ K1 , m) for each m,
n ≤ m ≤ 2n.

Theorem 1. For any graph G of order n, and every m, n ≤ m ≤ 2n, we have d(G ◦ K1 , m) =
n


m−n
22n−m .
Hence D(G ◦ K1 , x) = xn (x + 2)n .
Proof. Suppose that D is a dominating set of G ◦ K1 of size m, and |D ∩ V (G)| = i, for 0 ≤ i ≤ n.
Without loss of generality, suppose that V (G)∩D = {v1 , . . . , vi }. Therefore D contains ui+1 , . . . , un . For
 {v1 ,. . . , vi , ui+1 , ui+2 , . . . , un } to a dominating set D of size m with V (G)∩ D = {v1 , . . . , vi },
extending
i
we have m−n
possibilities. Therefore

n   
X n i
d(G ◦ K1 , m) = .
i =0
i m−n
n


It is not hard to see that the above sum is equal to m−n
22n−m . 
1716 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724

We recall the following lemma:

Lemma 2 ([1]). For any graph of order n, d(G, n) = 1. If G has no isolated vertex, then d(G, n − 1) = n.

In our proofs, we need the following theorem.

Theorem 2 ([1]). If a graph G has m components G1 , . . . , Gm , then D(G, x) = D(G1 , x) · · · D(Gm , x).

Here, we consider the following lemma.

Lemma 3. Let G be a graph of order n with t vertices of degree 1 and r isolated vertices. If D(G, x) =
i=1 d(G, i)x is its domination polynomial, then the following hold:
P n i

(i) r = n − d(G, n − 1).

(ii) If G has s K2 -components, then d(G, n − 2) = − t + s − r (n − 1) + 2r .
n



2
(iii) If G has no isolated vertices and D(G, −2) 6= 0, then t = 2 − d(G, n − 2).
n


(iv) d(G, 1) = |{v ∈ V (G)|deg (v) = n − 1}|.

Proof. (i) Suppose that A ⊆ V (G) is the set of all isolated vertices. For any vertex v ∈ V (G) \ A, the
set V (G) \ {v} is a dominating set of G. Therefore d(G, n − 1) = |V (G \ A)| = n − r.
(ii) Suppose that D ⊆ V (G) is a set of cardinality n − 2 which is not a dominating set of G. We have
two cases for D:
Case 1. D = V (G)\{v, w}, where v is an isolated vertex and w ∈ V (G)\{v}. Thus for every isolated
vertex v , there are n − 1 vertices such that V (G) \ {v, w} is not a dominating set. Therefore, the
total number of (n − 2)-subsets of V (G) of the form V (G) \ {v, w} which is not a dominating set
(v or w is an isolated vertex) is r (n − 1) − 2 , since if v and w are isolated vertices, then we
r


count V (G) \ {v, w} for both v and w .
Case 2. D = V (G) \ {v, w} for two adjacent vertices v and w with deg (v) = 1. Since we have s
K2 -components, the number of such {u, w} is t − s and the proof is complete.
(iii) Since D(G, −2) 6= 0, by Theorem 2, G has no K2 -component, and so by Part (ii), we obtain the
result.
(iv) For every v ∈ V (G), {v} is a dominating set if and only if v is adjacent to all vertices. The proof is
complete. 

Corollary 1. If the domination polynomial of a graph G is xn (x + 2)n , then G has no isolated vertices.

Proof. By Lemma 3, Part (i), the number of isolated vertices, r = 2n − d(G, 2n − 1) = 2n − 2n = 0.

Therefore we have the result. 

We recall the following two theorems.

Theorem 3 (Ore [5], p.41). If G is a graph of order n with no isolated vertex, then γ (G) ≤ b 2n c. 

Theorem 4 (Payan and Xuong [6]). For a graph G with even order n and no isolated vertices, γ (G) = n
2
if
and only if the components of G are the cycles C4 or the coronas H ◦ K1 for some connected graph H. 

Now, we determine all graphs with domination polynomial xn (x + 2)n .

Theorem 5. Let G be a graph. Then D(G, x) = xn (x + 2)n if and only if G = H ◦ K1 for some graph H of
order n.

Proof. One side is a consequence of Theorem 1. For the other side let G be a graph with D(G, x) =
xn (x + 2)n . Thus |V (G)| = 2n. By Corollary 1, G has no isolated vertex. Since γ (G) = n, by
Theorem 4, every component of G is a cycle C4 or a corona L ◦ K1 for some connected graph L. Since
D(C4 , x) = x2 (x2 + 4x + 6) does not divide xn (x + 2)n , we conclude that there exists a graph H such
that G = H ◦ K1 and the proof is complete. 
 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724 1717

3. Graphs with two domination roots

In this section, we characterize all graphs with at most two domination roots.
First, we characterize graphs with exactly one domination root. Since 0 is a root of any domination
polynomial of graph G, by Lemma 3, Part (i), we conclude that a graph G has one domination root if
and only if G is a union of isolated vertices.
We now characterize all graphs with two domination roots.

Theorem 6. Let G be a connected graph with exactly two distinct domination roots. Then D(G, x) =
xn (x + 2)n , where n is a natural number. Indeed G = H ◦ K1 for some graph H of order n.

Proof. Since 0 is a domination root with multiplicity γ (G) for every graph G and D(G, x) has two
distinct roots, we have D(G, x) = xi (x + a)m−i for some i ∈ N and a > 0, where m = |V (G)|. Therefore
the coefficient of xm−1 is (m − i)a, and so (m − i)a ∈ N ∪ {0}. This means a is a rational number. Since
every rational algebraic integer is an integer, we have a ∈ N. Now, we show that a = 2 and i = m 2
.
Since G is connected, the coefficient of xm−1 is m. Hence, m = (m − i)a. But i 6= 0, so we have a ≥ 2. By
m(a−1)
Theorem 3, i = a
= γ (G) ≤ m2 . Thus we have a = 2 and i = m2 . Therefore D(G, x) = xn (x + 2)n
for some n ∈ N. By Theorem 5, G = H ◦ K1 , for some graph H and the proof is complete. 

graph with V (H ) = {v1 , . . . , vn }. By Theorems 1 and 2, for any graph G of the

Let H be an arbitrary Q
form H ◦ K1 , D(G, x) = i=1 D(K2 , x). By an argument similar to that in the proof of Theorem 1, we
n

have the following theorem:

Theorem 7. Let Hi , 1 ≤ i ≤ n, be a graph containing a vertex vi of degree |V (Hi )| − 1, |V (Hi )| ≥ 2, and

H be a graph with vertex set V (H ) = {u1 , .Q
. . , un }. If G is the graph formed by identifying the vertex vi
with ui for every 1 ≤ i ≤ n, then D(G, x) = i=1 D(Hi , x). In other words, the domination polynomial of
n

such graph does not depend on the geometrical structure of H.

4. Graphs with three domination roots

In this section, we shall characterize all graphs with exactly three distinct domination
√
roots. First,
in the following theorem, we characterize all graphs with Z (D(G, x)) = {0, −3±2 5 }.
√
Theorem 8. Let G be a connected graph of order n. Then, Z (D(G, x)) = {0, −3±2 5
} if and only if
n n
G = H ◦ K 2 , for some graph H. Indeed D(G, x) = x (x + 3x + 1) .
3 2 3

n
Proof. Suppose that G = H ◦ K 2 ; then by Theorem 7, D(H ◦ K 2 , x) = D(K n ◦ K 2 , x) = (D(P3 , x)) 3 =
√ 3
n n
x 3 (x2 + 3x + 1) 3 . Therefore Z (D(G, x)) = {0, −3±2 5
}.
√
Now, we prove the necessity. Since the minimal polynomial of −3±2 5 over Q is x2 + 3x + 1, we
have D(G, x) = xi (x2 + 3x + 1)j . We have i + 2j = n. On the other hand, since for every connected
n n
graph of order n, d(G, n − 1) = n, we have i = j = 3n . Therefore D(G, x) = x 3 (x2 + 3x + 1) 3 , and
so G has a unique γ -set. Note that for every connected graph H, if d(H , γ (H )) = 1, then no vertex
of degree 1 is contained in the γ -set. This is because one can replace the vertex of degree 1 with its
neighbor to obtain another γ -set. Assume that D ⊆ V (G) is the unique γ -set of G. We have |D| = 3n .
n n
n2
The coefficient of xn−2 in x 3 (x2 + 3x + 1) 3 is 2
− 7n 6
. On the other hand, by Lemma 3, this coefficient is
, and so for every v ∈ V (G) \ D,
n
Therefore, t = 2n


2
− t, where t is the number of vertices of degree 1. 3
we have deg (v) = 1. Since D is a dominating set for G, and G is connected, V (G) \ D is an independent
set. We claim that for every v ∈ D, N (v) ∩ (V (G) \ D) 6= ∅. To see this, suppose that v ∈ D and
N (v) ∩ (V (G) \ D) = ∅. Since G is connected, there exists a vertex w ∈ N (v) ∩ D. Now, clearly D \ {v} is
a dominating set for G, a contradiction, and the claim is proved. Suppose that |N (vi )∩(V (G)\ D)| = ni ,
ni ≥ 1, for i = 1, . . . , 3n , where D = {v1 , . . . , v n }. Since for every v ∈ V (G) \ D, deg (v) = 1, for every
3
1718 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724

1 ≤ i < j ≤ 3n , N (vi ) ∩ N (vj ) ∩ (V (G) \ D) = ∅. Since for every natural n, D(K1,n , x) = x(1 + x)n + xn
[1], by Theorems 2 and 7, we have
n n
3
Y 3
Y
D(G, x) = D(K1,ni , x) = x(1 + x)ni + xni .


i =1 i=1
n n n
On the other hand D(G, x) = x (x2 + 3x + 1) 3 = (x(1 + x)2 + x2 ) 3 . We shall prove that ni = 2, for
3

every i, 1 ≤ i ≤ 3n . We claim that for every k 6= 2, x(1 + x)k + xk and x(1 + x)2 + x2 have no non-
zero common root. Suppose that α 6= 0 is a common root of x(1 + x)k + xk and x(1 + x)2 + x2 . Then
(1 + α)2 = −α , and (1 + α)k = −α k−1 . These imply that (1 + α)2k = (−1)k α k and (1 + α)2k = α 2k−2 .
Thus we√have |α|k−2 = 1. Since k 6= 2, we have |α| = 1. But the non-zero roots of x(1 + x)2 + x2
are −3±2 5
whose absolute values are not 1. So the claim is proved. Hence for every i, 1 ≤ i ≤ n
3
, we
conclude that ni = 2. Therefore G = H ◦ K 2 , for some graph H. 

Theorem 9. Let G be a graph with no pendant vertex. If D(G, x) is the domination polynomial of G and it
has exactly three distinct roots, then
( √ )
√ −3 ± 3i
Z (D(G, x)) ⊆ 0, −2 ± 2i, .
2

Proof. Clearly, it suffices to prove the theorem for a connected graph. Let G be a connected graph of
order n. Let Z (D(G, x)) = {0, a, b}, where a 6= b. Thus D(G, x) = xi (x − a)j (x − b)k , for some i, j, k. By
Lemma 2, we have
− (ja + kb) = n. (1)
Also by Lemma 3, Part (ii), we have
   
j k n
a2 + b2 + jkab = d(G, n − 2) = . (2)
2 2 2
Let P (x) be the minimal polynomial of a over Q. Clearly, all roots of P (x) are simple. This implies
that deg (P (x)) = 1 or 2. We consider two cases.
Case 1. deg (P (x)) = 2. In this case since D(G, x) has three distinct roots, the minimal polynomial of
b over Q is also P (x). Thus we have D(G, x) = xi (x2 + rx + s)j , where P (x) = x2 + rx + s. We have
i + 2j = n, and also by (1), −j(a + b) = n. By Theorem 3, i ≤ 2n . Hence j ≥ 4n . Since −j(a + b) = n,
and a + b is an integer, we have −(a + b) ∈ {1, 2, 3, 4}. Therefore we have four cases to consider:
Subcase 1.1. If a + b = −1. Then j = n, a contradiction.
n
Subcase 1.2. If a + b = −2. Then j = 2
, a contradiction.
n n
Subcase 1.3. If a + b = −3. This implies that i = j = n
3
. Thus we have D(G, x) = x 3 (x2 + rx + s) 3 .
Now, by (2) we have
n
n2 ab n
3
( a2 + b 2 ) + = d(G, n − 2) = .
2 9 2

Noting that a + b = −3, we conclude that a2 + b2 = 9 − 2ab. Thus by simple calculation we obtain
nab = 3n. Therefore ab = 3. Now, using the fact that a + b = −3, we have
( √ √ )
−3 − 3i −3 + 3i
a∈ , .
2 2

Subcase 1.4. Now, assume that a + b = −4. Then j = 4n and so i = γ (G) = 2n . By Theorem 4, G = C4
or G = H ◦ K1 , for some graph H. But H ◦ K√
1 has exactly two distinct roots −2 and 0. Thus G = C4 .
Moreover, we have Z (D(C4 , x)) = {0, −2 ± 2i}.
 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724 1719

Case 2. deg (P (x)) = 1. So D(G, x) = xi (x − a)j (x − b)k , where −a, −b ∈ N. Note that i + j + k = n.
By our assumption and Theorem 4, i < 2n . So j + k > 2n . If a ≤ −2 and b ≤ −2, then −(ja + kb) > n,
  we can assume that a = −1. Thus, by (1), j − kb = n. On the other hand,
a contradiction. Therefore
j k n
. Therefore, we have j2 − j + (k2 − k)b2 − 2jkb = n2 − n.


by (2), we have 2
+ 2
b2 − jkb = 2

This yields that (j − kb)2 − j − kb2 = n2 − n. So we have j + kb2 = n. Since j − kb = n, then we have
b = −1, a contradiction. 

Using Maple, we found that −1 is the domination root of no graph with at most six vertices. We
think that this is true for all graphs.

Conjecture 1. For every graph G, D(G, −1) 6= 0.

Remark 1. If Conjecture 1 is true, then by an argument similar to that given in the proof of Theorem 9,
one can show that, for every graph G with three domination roots,
( √ √ )
−3 ± 5 √ −3 ± 3i
Z (D(G, x)) ⊆ −2, 0, , −2 ± 2i, .
2 2

Theorem 10 ([2]). Let G be a graph containing a simple path of length at least 3. Then

D(G, x) = x [D(G ∗ e1 , x) + D(G ∗ e1 ∗ e2 , x) + D(G ∗ e1 ∗ e2 ∗ e3 , x)]

where e1 , e2 and e3 are three edges of a simple path, G ∗ e is the graph obtained from G by contracting the
edge e, and G ∗ e1 ∗ e2 = (G ∗ e1 ) ∗ e2 and G ∗ e1 ∗ e2 ∗ e3 = ((G ∗ e1 ) ∗ e2 ) ∗ e3 .

Theorem 11. Paths and cycles do not have −1 as a domination root.

Proof. We prove the theorem for paths Pn ; the proof for cycles is similar. By induction on n, we show
that D(Pn , −1) is an odd integer. Since D(P1 , x) = x, D(P2 , x) = x2 + 2x and D(P3 , x) = x3 + 3x2 + x,
the result is true for P1 , P2 and P3 . Now, suppose that the result is true for all path Pi of order less than
n. By Theorem 10,

D(Pn , −1) = − [D(Pn−1 , −1) + D(Pn−2 , −1) + D(Pn−3 , −1)] .

Since the sum of three odd numbers is odd, we have the result. 

5. D -equivalence classes of some graphs

In this section, we investigate the D -equivalence classes of some graphs. First, we provide a
formula for the computation of the domination polynomial of the join of two graphs.

Theorem 12. Let G1 and G2 be graphs of orders n1 and n2 , respectively. Then

D(G1 ∨ G2 , x) = (1 + x)n1 − 1 (1 + x)n2 − 1 + D(G1 , x) + D(G2 , x).

Proof. Let i, 1 ≤ i ≤ n1 + n2 , be a natural number. We want to determine d(G1 ∨ G2 , i). If i1 and i2

are two natural numbers such that i1 + i2 = i, then clearly, for every D1 ⊆ V (G1 ) and D2 ⊆ V (G2 ), for
which |Dj | = ij , j = 1, 2, D1 ∪ D2 is a dominating set for G1 ∨ G2 . Moreover, if D ∈ D (G1 , i), then D is
a dominating set for G1 ∨ G2 of size i. The same is true for every D ∈ D (G2 , i). Thus

D(G1 ∨ G2 , x) = (1 + x)n1 − 1 (1 + x)n2 − 1 + D(G1 , x) + D(G2 , x). 

Corollary 2. Assume that G is a graph of order n and v ∈ V (G). If deg (v) = n − 1, then G is D -unique
if and only if G \ {v} is D -unique. Hence Kn and K1,n are D -unique for every natural number n.
1720 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724

Proof. By Theorem 12, D(G, x) = x (1 + x)n−1 − 1 + x + D(G \ {v}, x). Thus G is D -unique if and


only if G \ {v} is D -unique. 

We need the following lemma for the investigation of D -equivalence of some graphs.

Lemma 4. Let G be a graph of order n with domination polynomial D(G, x) =

n1 d(G, i)x . Ifd(o
G, j) =
Pn i
  i=

for some j, then δ(G) ≥ n − j. More precisely, δ(G) = n − l, where l = min j|d(G, j) = j
n n
j
, and
there are at least l−1 − d(G, l − 1) vertices of degree δ(G) in G. Furthermore, if for every two vertices
n

of degree δ(G), say u and v , we have N [u] 6= N [v], then there are exactly l−1 − d(G, l − 1) vertices of
n

degree δ(G).
 
Proof. Since d(G, j) = j , for any r, r ≥ j, every r-subset of the vertices of G forms a dominating
n

set for G. To obtain a contradiction suppose that there is a vertex v ∈ V (G) such that deg (v) < n − j.
Consider V (G) \ N [v]. Clearly, |V (G) \ N [v]| ≥ j, and V (G) \ N [v] is not a dominating set for G, a
contradiction. Now, assume that S is a (l − 1)-subset of V (G) which is not a dominating set. Thus there
is a vertex u ∈ V (G) \ S which is not covered by S. Since δ(G) ≥ n − l, we have deg (u) = n − l. Let
S 0 6= S be a (l − 1)-subset which is not a dominating set. Thus there exists a vertex u0 ∈ V (G)\ S 0 which
is not covered by S 0 . As we found before, deg (u0 ) = n − l. We claim that u 6= u0 . Since S 6= S 0 , there
exists a vertex x ∈ S ∩ (V (G) \ S 0 ). If x = u0 , then it is clear that u 6= u0 , because u 6= x. We know that
u0 x ∈ E (G). If u = u0 , then ux ∈ E (G), a contradiction. Thus u 6= u0 and the claim is proved. If u is any
vertex of degree δ(G), then S
= V (G) \ N [u] is an (l − 1)-subset of V (G), which is not a dominating set.
Hence, there are at least l−1 − d(G, l − 1) vertices of degree δ(G). Now, suppose v ∈ V (G) \ S = N [u]
n

and that v has degree δ(G) and is not adjacent to any vertex in S. Then N [v] = V (G) \ S = N [u]. So, if
N [x] 6= N [y] for every two vertices x and y of degree δ(G), then the vertex
u is unique, and there is a
bijection between the sets S and the vertices u. Thus there are exactly l−1 − d(G, l − 1) vertices of
n

degree δ(G). 

Corollary 3. Let G be a graph such that D(G, x) = D(Kn,n , x). Then G is an n-regular graph.
 
Proof. By Theorem 12, D(Kn,n , x) = ((1 + x)n − 1)2 + 2xn . Thus d(G, n) =
2n
n
. Hence by Lemma 4,
G has 2n vertices of degree n, as desired. 

Remark 2. The D -equivalence class of Kn,n is not unique. Consider two disjoint copies of the complete
graph Kn with vertex sets {v1 , . . . , vn } and {v10 , . . . , vn0 }, respectively. Join vi to vi0 , for every i, 1 ≤ i ≤ n.
It can be easily seen that the domination polynomial of this graph is the same as the domination
polynomial of Kn,n . Indeed, by Lemma 4, we have the following theorem which gives the relationship
between the domination polynomial of a graph G and the regularity of G.

Theorem 13. Let H be a k-regular graph with N [u] 6= N [v] for every u, v ∈ V (H ). If D(G, x) = D(H , x),
then G is also a k-regular graph.

As an application of Theorem 13, we shall investigate the equivalence classes of cycles. We need
the following theorem:

(i) For every n ≥ 3, d(Pn , n − 2) =

n


Theorem 14. − 2 [3].
2
(ii) For every n ≥ 4, d(Pn , n − 3) = ( )
n


3
− 3n − 8 [3].
(n+2)(n+3)
(iii) For every n ∈ N, d(P3n , n) = 1, d(P3n+2 , n + 1) = n + 2 and d(P3n+1 , n + 1) = 2
− 2 [3].
(iv) If n = 3k, k ≥ 1, then γ (Cn ) = k, and d(Cn , k) = 3 [2].
2 5n
(v) If n = 3k + 1, k ≥ 1, then γ (Cn ) = k + 1, and d(Cn , k + 1) = n +6
[2].
(vi) If n = 3k + 2, k ≥ 1, then γ (Cn ) = k + 1, and d(Cn , k + 1) = n [2].

Theorem 15. Let n be a natural number. If n ≡ 0, 2 (mod 3), then Cn is D -unique.

 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724 1721

Proof. Assume that G is a graph such that D(G, x) = D(Cn , x). By Theorem 13, G is 2-regular graph.
Hence G is a union of finitely many disjoint cycles. Let G be the disjoint union of cycles Cm1 , . . . , Cmr .
First suppose that n ≡ 0(mod 3). Since D(G, x) = i=1 D(Cmi , x) = D(Cn , x) and d(Cn , 3 ) = 3, by
Qr n

Theorem 14, we conclude that r = 1 and G = Cn .

Now, suppose that n ≡ 2(mod 3). If there exists some j, such that mj ≡ 0(mod 3), noting that
m
d(Cn , d 3n e) = n and d(Cmj , 3j ) = 3, we have n ≡ 0(mod 3), a contradiction. Thus for each i,
n2 +5n
1 ≤ i ≤ k, mi ≡ 1 or 2(mod 3). Clearly, for every natural number n, n ≤ 6
. This implies that
n
the coefficient of xd 3 e in the polynomial D(G, x) = i=1 D(Cmi , x) is at
Qr rQ
least i=1 mi . By induction
on r it is not hard to see that, if r ≥ 2 and m 1 , . . . , mr are natural numbers more than 2, then
i=1 mi > i=1 mi = n = d(G, d 3 e), a contradiction. Therefore r = 1 and G = Cn . The proof is
Qr Pr n

complete. 

Remark 3. We think that Cn is also D -unique for n ≡ 1(mod 3), but we are not able to prove it.

We now study the D -equivalence classes of paths. First, we need some results.

Lemma 5. Let G be a graph and u be a pendant vertex of G. Suppose that uv ∈ E (G), deg (v) = 2 and
N (v) = {u, w}. If G has a unique γ -set, then G \ {u, v, w} has a unique γ -set.

Proof. Assume that γ (G) = t, and D ⊆ V (G) is a unique γ -set of G of size t. Since u is a pendant vertex,
one of the vertices u and v should be contained in D. Obviously, v ∈ D, since otherwise (D \{u})∪{v} is
another γ -set for G, a contradiction. Clearly, w 6∈ D, since if w ∈ D, then (D \{v})∪{u} is another γ -set
for G, a contradiction. Hence |D ∩ (V (G) \ {u, v, w})| = t − 1. We claim that γ (G \ {u, v, w}) = t − 1.
Because, if γ (G \ {u, v, w}) < t − 1, then by adding v to a γ -set for G \ {u, v, w} we obtain a γ -set
for G with size at most t − 1, a contradiction. Thus γ (G \ {u, v, w}) = t − 1. For every γ -set S of
G \ {u, v, w}, S ∪ {v} is a γ -set of G. This implies that G \ {u, v, w} has a unique γ -set and the proof is
complete. 

It is easy to verify directly the following lemma.

Lemma 6. If n ≡ 0(mod 3), then Pn has a unique γ -set.

Now, we state the following lemma.

Lemma 7. Let G be a graph, u, v ∈ V (G) and deg (u) = deg (v) = 1. If uw, vw 0 ∈ E (G) and ww 0 6∈ E (G),
then D(G, x) = D(G + ww 0 , x), and hence G is not D -unique.

Proof. Clearly, every dominating set for G is a dominating set for G + ww 0 . Now, let S ⊆ V (G) be a
dominating set for G + ww 0 . If both w, w 0 ∈ S or both w, w 0 6∈ S, then obviously S is also a dominating
set for G. So suppose that w ∈ S and w 0 6∈ S (or w 6∈ S and w 0 ∈ S). Since S is a dominating set for
G + ww 0 , we have v ∈ S. This implies that S is a dominating set for G. Therefore we conclude that
D(G, x) = D(G + ww 0 , x) and the proof is complete. 

Lemma 8. (i) Let n ≥ 3 be a natural number and n ≡ 1 or 2 (mod 3). Then for every v ∈ V (Cn ), Cn has
at least two γ -sets containing v .
(ii) Let n ≥ 7 be a natural number and n ≡ 1(mod 3). If u, v ∈ V (Cn ) and uv 6∈ E (Cn ), then there are at
least two γ -sets of Cn containing u and v .

Proof. (i) If n ≡ 1 or 2 (mod 3), then by Theorem 14, Cn contains at least three γ -sets, because
2
n ≤ n + 6
5n
. Since γ (Cn ) = d 3n e, we conclude that at least two γ -sets of Cn have non-empty
intersection. This completes the proof.
(ii) We apply induction on n. It can be easily checked that the result is true for n = 7. Suppose that
n > 7 and n ≡ 1(mod 3). Assume that x, y, z ∈ V (Cn ) \ {u, v} and N (x) = {a, y}, N (y) =
{x, z }, N (z ) = {y, b}.
1722 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724

Fig. 1. A path with N (x) = {y, z }, N (y) = {x, a} and N (z ) = {x, b}.

Fig. 2. Two graphs of [Pn ].

Remove x, y and z and join a and b. By the induction hypothesis Cn−3 has at least two γ -sets containing
u and v . Let D be a γ -set for Cn−3 . If a ∈ D, then D ∪ {z } is a γ -set for Cn . If b ∈ D, then D ∪ {x} is a
γ -set for Cn . If a, b 6∈ D, then D ∪ {y} is a γ -set for Cn . Thus Cn has at least two γ -sets containing u and
v , and the proof is complete. 
We also need the following theorem to obtain the result on D -equivalence classes of paths.

Theorem 16. Let n ≥ 2 be a natural number. The size of the smallest dominating set containing both end
vertices of Pn is d n+
3
2
e. Moreover, if n ≥ 4 and n 6≡ 1(mod 3), there are at least two dominating sets of
size d n+
3
2
e containing both end vertices of Pn .
Proof. We apply induction on n. For n = 2, 3, 4 the assertion is trivial. Assume that x ∈ V (Pn ),
deg (x) = 2, N (x) = {y, z } and deg (y) = deg (z ) = 2. Let N (y) = {x, a} and N (z ) = {x, b}; see Fig. 1.
Remove x, y, z and join a and b. By the induction hypothesis, the size of the smallest dominating set
containing end vertices of Pn−3 is d n−3
1
e. If D is a dominating set for Pn−3 of size d n−3 1 e containing end
vertices and a ∈ D, then D ∪ {z } is a dominating set for Pn of size d n+
3
2
e. If a, b 6∈ D, then D ∪ {x} is the
desired dominating set for Pn . Now, suppose that Pn has a dominating set, say D, of size less than d n+3
2
e
containing both end vertices of Pn . It is not hard to see that there exists a v ∈ D such that deg (v) = 2
and N (v) ∩ D = ∅. Consider D \ N [v] and join two pendant vertices of two components of D \ N [v]
to obtain a path of order n − 3. This path has a dominating set of size less than d n− 3
1
e containing end
vertices of Pn−3 , a contradiction. Now, assume that n 6≡ 1(mod 3). For n = 5 and n = 6 the assertion
is trivial. Now, by induction on n and using a similar method to the one that we used before, the proof
is complete. 

Theorem 17. Let n be a natural number and n ≡ 0 (mod 3). Then [Pn ] contains just two graphs in Fig. 2.

Proof. Let G be a graph such that D(G, x) = D(Pn , x). By Lemma 3, Part (i), G has no isolated vertex. By
Theorem 14, d(G, 3n ) = 1. If D(G, −2) = 0, then 2|d(G, 3n ) = 1, a contradiction. Thus D(G, −2) 6= 0
and so by Lemma 3, Part (iii), the number of pendant vertices of G is 2. Note that since
G has no K2 -
component, two pendant vertices are not adjacent. By Theorem 14, d(Pn , n − 3) = 3 − (3n − 8).
n

This means that the number of (n − 3)-subsets which are not dominating sets of G is exactly 3n − 8.
We claim that the number of (n − 3)-subsets S which are not dominating sets and for which V (G) \ S
does not contain a pendant vertex is at least n − 4. For every pendant vertex v with neighbor w , the
complement of {v, w, x}, where x is an arbitrary vertex, is an (n − 3)-subset which is not a dominating
set. Since x has n − 2 possibilities and there are two pendant vertices, we have at most 2(n − 2) of
these subsets (note that some of these sets may be equal). Therefore the number of (n − 3)-subsets
S which are not dominating sets and for which V (G) \ S does not contain a pendant vertex is at least
3n − 8 − (2n − 4) = n − 4.
Now, suppose that S is not a dominating set, and V (G) \ S = {u, v, w}, and none of the vertices
u, v and w is pendant. Since S is not a dominating set of G, at least one of the vertices u, v and w
has no neighbor in S and so it is of degree 2. Without loss of generality assume that deg (u) = 2
and N (u) = {v, w}. Let S 0 ⊆ V (G) be an (n − 3)-subset which is not a dominating set, where
V (G) \ S 0 = {u0 , v 0 , w 0 } and none of the u0 , v 0 , w 0 is pendant. By the same argument we have
deg (u0 ) = 2 and N (u0 ) = {v 0 , w 0 }. If u = u0 , then N (u) = N (u0 ) and so S = S 0 .
 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724 1723

a b

Fig. 3. Two possibilities for G with D(G, x) = D(Pn , x).

Therefore, by the above argument, G has at least n − 4 vertices of degree 2 which are not adjacent
to a vertex of degree 1. Thus if G has exactly n − 4 vertices of degree 2, then no pendant vertex is
adjacent to a vertex of degree 2. Moreover, no pair of pendant vertices of G have a common neighbor.
In this case G is the graph shown in Fig. 3(a). If G has exactly n − 3 vertices of degree 2, then one of the
pendant vertices cannot be adjacent to a vertex of degree 2. So we have the graph shown in Fig. 3(b).
Note that in Fig. 3(a), Pn1 , . . . , Pnk (k ≥ 1) don’t include vertices u and v , and s ≥ 0 or r ≥ 0, and if
r = 0 or s = 0, then k ≥ 2. If s = 0, k = 1, or r = 0, k = 1, then we have Fig. 3(b).
Now, we consider the following cases:
Case 1. Suppose that D(G, x) = D(Pn , x) and G is the graph (a) shown in Fig. 3. Since D(G, x) = D(Pn , x),
G has a unique γ -set which should contain u and v . If there exists some j such that mj 6≡ 0(mod 3) or
rj 6≡ 0(mod 3), then by Lemma 8, Part (i), the γ -set of G is not unique, a contradiction. First assume
that for every j, nj ≥ 3. By Theorem 16, since G has a unique γ -set, nj + 2 ≡ 1(mod 3). Thus we have
k r s
X nj − 2 X mj X rj
γ (G) = +2+ −r + −s
i=1
3 j =1
3 j =1
3
k
P r
P s
P
nj + mj + rj
j =1 j =1 j =1 2k
= − − r − s + 2.
3 3
On the other hand,
k
X r
X s
X
n = |V (G)| = nj + 2 + 2 + (mj − 1) + (rj − 1)
j=1 j =1 j =1

k
X r
X s
X
= nj + mj + rj − r − s + 4.
j =1 j =1 j =1

Since γ (G) = 3n , we have k + r + s = 1. Since k ≥ 1, hence k = 1, s = 0, r = 0 and G is Pn . If there

exists some j such that 1 ≤ nj ≤ 2, then u and v dominate vertices of Pnj , and we have γ (G) < 3n , a
contradiction. Now, suppose that for some j, nj = 0. Without loss of generality we can assume that
n1 = 0. Since G has no multiple edge, for every j, 2 ≤ j ≤ n, nj ≥ 3. Thus we have
s
P
k r
rj
X nj − 2 X mj j =1
γ (G) = +2+ −r + −s
j =2
3 j =1
3 3
1724 S. Akbari et al. / European Journal of Combinatorics 31 (2010) 1714–1724

a b

Fig. 4. Two graphs for the case n1 = 0.

k
P r
P s
P
nj + mj + rj
j =2 j=1 j =1 2(k − 1)
= −r −s+2− .
3 3
Since γ (G) = n
3
, we conclude that k + r + s = 2. There are two cases to be considered:
Subcase 1.1. k = 2, r = s = 0. In this case, G is the graph (a) shown in Fig. 4. By Lemma 7,
D(G, x) = D(Pn , x).
Subcase 1.2. k = 1, r = 1, s = 0. In this case, G is the graph (b) shown in Fig. 4.
If D is the unique γ -set of G, then D contains u and v . Clearly (D \ {v}) ∪ {y} is another γ -set for G,
a contradiction.
Case 2. Let G be the graph (b) in Fig. 3.
Clearly, u is contained in the γ -set of G. If there exists some j such that mj 6≡ 0(mod 3), then by
Lemma 8, Part (i), we get a contradiction. We have
  r
n1 − 1 X mj
γ (G) ≤ +1+ − r.
3 i=1
3

On the other hand |V (G)| = n1 + 2 + mi − r. But if r > 1, then

Pr
i=1
  r r
n1 − 1 X mi n1 + 2 − r X mi
+ +1−r < + ,
3 i=1
3 3 i=1
3

which is a contradiction. If r = 1, then noting that n = |V (G)| is divisible by 3, we get n1 ≡ 2(mod 3).
But then the number of γ -sets of Pn1 +1 containing an end vertex is at least 2. Thus G has at least two
γ -sets, a contradiction. 

Acknowledgements

The authors would like to express their gratitude to the referee for her/his careful reading and
helpful comments. The first author is indebted to the Institute for Mathematical Research (INSPEM)
at University Putra Malaysia (UPM) for partial support and hospitality during his visit. The research of
the first author was in part supported by a grant from IPM (No. 88050212).

References

[1] S. Alikhani, Y.H. Peng, Introduction to domination polynomial of a graph, Ars Combin. (in press).
[2] S. Alikhani, Y.H. Peng, Dominating sets and domination polynomial of cycles, Glob. J. Pure Appl. Math. 4 (2) (2008) 151–162.
[3] S. Alikhani, Y.H. Peng, Dominating sets and domination polynomials of paths, Int. J. Math. Math. Sci., 2009, Article ID 542040.
[4] R. Frucht, F. Harary, On the corona of two graphs, Aequationes Math. 4 (1970) 322–324.
[5] T.W. Haynes, S.T. Hedetniemi, P.J. Slater, Fundamentals of Domination in Graphs, Marcel Dekker, New York, 1998.
[6] C. Payan, N.H. Xuong, Domination-balanced graphs, J. Graph Theory 6 (1982) 23–32.